eipedr], and c«rei for elirrj. Tribal names like Oivyp, -vjJSos (not Oivqis,
-vrjiSos, as commonly edited), are equally spelt -ei's, -eiSos. Hetero-
clitic datives from proper nouns in -17s occur, as YlturiOd&a for -S77.1
The inscriptions of various dialects give indications that the /-element
was weaker and sooner lost in the case of 771 than of the other like
diphthongs. Moreover, the pronunciation of 771 was facilitated by
the shortening of ?; before 1; thus it became a simple closed ?-sound
not widely different from e.2 Indeed, E can be written for 771; thus
XoAk[o^i/]k£ airi (-#77/07 avrfj, II. 61, c. 358). Conversely, prjvvo-h
(nom. pi.) is spelt nrjvvo-qis in 572 (circa Eucl. anh:).8 In short, we
have examples of u put for 771, 771 for a and for e, and e for r/t. The
two diphthongs a and 771 had become almost indistinguishable from
each other and from c. Two important inferences can be drawn
from this : first, that the phonetic decay of u did not, as has been gen-
erally assumed (and as was the case in Latin), consist in a gradually
increasing preponderance of its /-element, but in a gradual loss of the
distinctness of each separate element, so that ei became a close e not
more diphthongal than I. Second, that this change had been nearly,
if not quite, consummated before the end of the Attic period.4
1 Wecklein, Cur. Epigr., p. 63; Herwerden, Lap. Test., pp. 5, 6. Perhaps
the confusion of datives hastened that of accusatives, and finally that of genitives.
By the beginning of the third century the false analogy has given the declension
of a-stems to s-stems of proper nouns throughout. Perhaps in this, as in many
things, Xenophon's usage anticipated the common dialect, and we should, with
the manuscripts, give him such forms as %aKpirr\v for Attic ZaKparv, etc.
2 The shortening of 77 before 1 (cf. Dittenberger in Hermes, XVII. 37) is
confirmed by the analogy of other diphthongs; ydvs, e.g., became mis, as other-
wise the Attic form must have remained vnus. (G. Meyer, Gr. 118, 298.) Cf.
Kw/ioiS!a for KuiifSla, Kaibel, Ep. Gr. 38.
8 Other examples are : ir6\ri, II. 25 (before 376?); $ov\e7, 38; n-pax6e[t],
S6(ei (subj.), 49; Olyels, 55; xoXko0>Vc«, four times, beside -07)Ke, 61; i[i>oi]x8ei,
irapcL<TK[eva]<T8ti, id.; [-ypap.p.~\ao-ri, e[I] for ;T, 90; re?, avret, etc.; -riu.l\aei, o-rt-
<pavd<rei (subj.); so Sonet, twice; eipeBri, 114; [fl"«> 125; o-ri)Kei KtBivet, 147;
<rvvTe\ea6e?, irapaKaBti, ■KTwp.tniaei, 167; A'ryeTSos, 16S (and 'AptrrTelSys, etc., for
Tl'iS-qs after 400) ; rpiaKoare'i, 1S0; eopreT, 577; el, passim (= ?J) ; e&Sofiei, evaret,
id.; ayaBe! rvxei, re7 tpv\eT, 564; eKrei nal Seicarei, 175 /' (Add. et Corrig.); a<j>ie7,
573^ (>d.). The list may be easily extended by reference to the C. I. A., II.
These citations are enough to show that analogy, in this case, has nothing to do
with the changed spelling.
4 In the fourth century, 771 is found (in C. I. A., II. I; v. Meisterh., Gr. p. 18)